On Tue, Mar 24, 2015 at 12:52 AM, John Clark <[email protected]> wrote:
> On Mon, Mar 23, 2015 at Telmo Menezes <[email protected]> wrote: > > > In my opinion the fundamental problem with the Turing Test is that >> passing it is an act of deception. The computer has to fake being a human. >> > > Lying takes intelligence, some have even suggested that the ability to > deceive our fellows was of the driving evolutionary forces that drove the > increase in brain size, but if you have scruples about it then how about > this; you are free to ask any question except "are you a machine?" that > way it can always tell the truth. > Not really. Most topics that humans chose when trying to make a connection with another human would require the machine to lie. "how old are you?", "what do you do for a living?", "are you a man or a woman?", "are you married? / why not?". The list of questions you would have to ban is immense, and still I'm sure it would be possible to find a loophole. > And anyway the really important thing isn't if you can detect if the thing > you're talking to is a human but if you can detect if the thing you're > talking to is intelligent. > Yes, that is the important thing. That is not what the Turing Test asks, though. It tries to demonstrate intelligence indirectly, and I am arguing that it is possible to pass statistical significance while not developing the sort of technology that Turing and all the other AI dreamers have in mind. In fact, this has been accomplished recently and still we don't have HAL 9000. > > >> > the computer has no human body, >> > > That's just cosmetic. > Far from it. The constraints and possibilites of the human body shape our intelligence to a large degree. I could give you many examples. "I felt butterflies in my stomach"; the decimal system because we have ten fingers; the very characters the computer uses to communicate with you are a representation of sounds made by our specific vocal chords, tongue and mouth, etc etc etc. I fully believe it is possible to have an advanced intelligence without a human body, but it will not be a human, which is what the Turing Test tests for. > As far back as the 1964 world's fair Disney made a pretty convincing > animated manikin of Abraham Lincoln, but I don't find that very > interesting. I've never met you so I don't know what you look like and > don't much care, but if I found out you didn't look like anything at all > because you were a computer program I would find that very interesting > indeed. > > > Human behavior is full of patterns, that can be exploited by brute >> force. This is what Watson does, essentially. Watson is more or less a >> traditional database of character strings with sophisticated indexing and >> querying algorithms. >> > > I confess that I don't have much patience when people say yes the machine > behaved very intelligently, more intelligently than I did, but it doesn't > count as being *really* intelligent because the machine was only successful > in solving that very difficult problem because it did it its way rather > than my inferior way that didn't work. > I don't question that Watson and Deep Blue and the Mars Rover display intelligence. What I claim is that it is a much narrower form if intelligence than human intelligence. Narrower intelligence can be better than us at specific tasks while being far from our ability to generalise. Even language is interpreted in a very narrow sense with Watson. It can essentially discover the most plausible answer to a question (or the most plausible question to an answer) by doing some very sophisticated pattern matching against a very large corpus of texts generated by humans. It cannot invent a joke or write a novel. I believe there will be AIs capable of inventing jokes and writing novels, but I also believe they will require a vastly different algorithm. Like you, I really look froward to that algorithm being found, so exaggerations about the capabilities of Watson or the effectiveness of the Turing Test annoy me, because I feel they start being in the way of that goal. > > > >> > Watson appears to be an amazing piece of software and I think it >> displays intelligence, but in a much narrower fashion than the hype >> surrounding it seem to assume. >> > > Deep Blue was the world chess champion but it's intelligence was very > narrow, however on Jeopardy Watson's demonstration that he understood the > questions, never mind that he knew the answers, showed that his > intelligence was much broader. > They are quite similar actually (a gigantic search tree, pruned and guided by heuristics). Watson became possible later because it requires for most areas of human knowledge to be covered by texts available for easy retrieval through some digital medium. This only happened very recently. I have nothing against this sort of effort, I've been working in that field myself. This does not mean that gigantic search trees cannot be used to create generic intelligence, but I am quite skeptical that they can. This skepticism comes from the observation that the human brain is highly associative, distributed and asynchronous. So I tend to think these will be properties of the generic algorithm. I agree with you that some people will be always bigoted towards machine intelligence, but that is not what I'm saying at all. > > > I propose a different test. I show you a computer program that you can >> have a conversation with. You talk with it for half an hour and then I tell >> you I'm going to shut it down forever. It will essentially die. How >> distressed are you? What if I point a gun at a bonobo monkey? >> > > That would just test for what somebody thought was cute not what they > thought was intelligent. People are suckers for anything with juvenile > characteristics like a big head, small jaw, receding chin, bulging cranium > and pudgy legs. Knock together a robot that has all that and it would be > cute as a button even if it was dumb as dog shit. > That is true (I believe I admitted it myself when writing the example). So this is why I sent you that video. Did you watch it? Those robots have no juvenile characteristics nor are they cute in the traditional sense. In fact they are military weapons. A lot of people on the Internet have been complaining about that guy kicking the robot. My claim is that this is because the robot displays a lot if intelligence in its movements. The AI of these robots is a secret, but I would bet it is driven by some sophisticated neural networks, something much closer to what I mentioned above. It is not pretending to be something it is not, nor pretending to understand words it does not. Telmo. > > John K Clark > > > > > > > > -- > You received this message because you are subscribed to the Google Groups > "Everything List" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at http://groups.google.com/group/everything-list. > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "Everything List" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/everything-list. For more options, visit https://groups.google.com/d/optout.
